SKILL.md
489 lines · 1,510 words
ML Pipeline
Overview
A production ML pipeline transforms raw data into deployed predictions through a series of reproducible, automated stages. This skill covers end-to-end pipeline design from data ingestion through model training, scoring, versioning, and orchestration, with emphasis on reproducibility and experiment tracking.
Pipeline Architecture
+----------+ +-----------+ +----------+ +----------+ +---------+
| Data | -> | Feature | -> | Model | -> | Model | -> | Model |
| Ingest | | Engineer | | Training | | Scoring | | Registry|
+----------+ +-----------+ +----------+ +----------+ +---------+
| | | | |
v v v v v
[DVC] [Feature [Experiment [Metrics [MLflow
Data Store] Tracker] Store] Registry]
Version